Direct Comparison of In Vivo Fate of Second and Third-Generation CD19-Specific Chimeric Antigen Receptor (CAR)-T Cells in Patients with B-Cell Lymphoma: Reversal of Toxicity from Tonic Signaling

Although adoptive transfer of T cells with second-generation CD19-specific CARs containing CD28 or 4-1BB costimulatory endodomains shows remarkable clinical efficacy against B cell malignancies, the optimal choice of costimulatory domains in these and other CARs remains controversial. Depending on the precise CAR structure and specificity, individual endodomains may be associated with deleterious ligand-independent tonic signaling in the transduced T cell. Long et al. (Nat Med 2015) established the CD28 co-stimulatory endodomain can have a toxic tonic signaling effect, but it is unclear if tonic 4-1BB signaling may have deleterious consequences as well, and if such effects can be reversed.

We therefore modeled tonic CAR signaling in T cells by transducing them with gammaretroviral vectors expressing 2^nd-generation CD19.CAR constructs containing either the CD28 or 4-1BB costimulatory endodomain (in addition to the CD3-ζ chain endodomain). Compared to CAR-T cells with the CD28 endodomain alone, those with 4-1BB alone expanded 70% more slowly following transduction. Impaired expansion of 4-1BB CD19.CAR-T cells was coupled with a 4-fold increase in apoptosis and a gradual downregulation of CAR expression, and was a consequence of 4-1BB-associated tonic TRAF2-dependent signaling, leading to activation of NF-κB, upregulation of Fas and augmented Fas-dependent activation-induced T cell death (AICD). Moreover, expression of 4-1BB CAR from a gammaretroviral vector increased tonic signaling through a self-amplifying/positive feedback effect on the retroviral LTR promoter.

Because of the toxicity of 4-1BB in our gammaretroviral CAR.CD19 construct (manifest by delayed expansion and increased apoptosis) we could not directly compare the in vivo fate of T cells expressing CAR.CD19 4-1BB with that of co-administered CAR.CD19 CD28 T cells in patients with lymphoma. We found, however, that the adverse effects of tonic 4-1BB costimulation could be overcome in a 3^rd-generation CAR.CD19 vector, containing both CD28 and 4-1BB costimulatory molecules in tandem.

We thus compared the fate of a 3^rd-generation vector containing both CD28 and 4-1BB costimulatory domains with that of a 2^nd-generation vector containing CD28 alone. Six patients with refractory/relapsed diffuse large B-cell lymphoma received 2 cell populations, one expressing 2^nd and one expressing 3^rd generation vectors. To determine whether CD28 alone was optimal (which would suggest 4-1BB is antagonistic) or whether 4-1BB had an additive or synergistic effect contributing to superior persistence and expansion of the CD28-41BB combination, patients were simultaneously infused with 1-20×10⁶ of both 2^nd and 3^rd generation CAR⁺ T cells/m² 48-72 hours after lymphodepletion with cyclophosphamide (500 mg/m²/d) and fludarabine (30 mg/m²/d) × 3. Persistence of infused T cells was assessed in blood by CD19.CAR qPCR assays specific for each CAR. Molecular signals peaked approximately 2 weeks post infusion, remaining detectable for up to 6 months. The 3^rd-generation CAR-T cells had a mean 23-fold (range 1.1 to 109-fold) higher expansion than 2^nd-generation CAR-T cells and correspondingly longer persistence. Two patients had grade 2 cytokine release syndrome, with elevation of proinflammatory cytokines, including IL-6, at the time of peak expansion of T cells. Of the 5 patients evaluable for response, 2 entered complete remission (the longest ongoing for 9 months), 1 has had continued complete remission after autologous stem cell transplantation, 1 had a partial response, and 1 progressed.

In conclusion, our data indicate that infusion of T cells carrying a CD19.CAR containing CD28 and 4-1BB endodomains is safe and can have efficacy at every dose level tested. Additionally, in a side-by-side comparison, the 3^rdgeneration vector produced greater in vivo expansion and persistence than an otherwise identical CAR-T cell population with CD28 alone.